What problem does it solve?
Agent sessions and .squad state can consume large amounts of the model context window over time, causing higher token costs, degraded performance, and reduced ability for agents to reason across long-running workflows. The nap skill reduces context bloat by compressing histories, pruning old logs, archiving stale decisions, and cleaning orphaned inbox files so active agents work with a smaller, relevant context set.
Core Features & Use Cases
- Compress histories: Reduce token footprint by compressing or summarizing agent history files when they grow large.
- Prune and archive: Remove or archive old logs and stale decision records to persistent storage to keep working directories lean.
- Orphan cleanup & modes: Detect and remove orphaned inbox files and support dry-run and deep-clean modes for safe maintenance.
- Use Case: Run before a large fan-out of agents or after prolonged sessions to ensure new agents start with a compact and relevant context window.
Quick Start
Run the nap cleanup to compress histories, prune old logs, archive stale decisions, and remove orphaned inbox files before starting a large fan-out.